McNemar's test is an ordinary approximation test which evaluates the significance of the variation between two correlated proportions, where the two proportions are based on the same sample of subjects or on matched-pair samples.
A researcher tries to find out if a particular drug has an effect on a specific disease. Counts of individuals are given in the below table, with the diagnosis (disease: present or absent) before treatment given in the rows, and the diagnosis after treatment in the columns. The test needs same subjects to be added in the before-and-after measurements.
Disease:present |
Disease:absent |
Row total | |
Disease:present |
a:50 | b:60 | 110 |
Disease:absent |
c:40 | d:30 | 70 |
Column total | 90 | 90 | 180 |
a=50 b=60 c=40 d=30
McNemar test statistic McNemar Chi Square test statistic(Χ2) Difference Between Test Odds Radio
Substitute the given values in the formula of Χ2,
Χ2 = (b - c)2 / (b + c) Χ2 = (60 - 40)2 / (60 + 40) Χ2 = (20)2 / 100 Χ2 = 400 / 100 = 4
Statistic with Yates's correction for continuity calculation: Substitute the given values in the formula, Χ2= (|b - c| - 0.5)2 / (b + c) Χ2= (|60 - 40| - 0.5)2 / (60 + 40) Χ2= (20 - 0.5)2 / 100 Χ2= 380.25 / 100 Χ2= 3.8025
Difference Between Test calculation: Substitute the given values of a,b,c and n in the formula, DF= (a+b) / n - (a+c) / n n = a+b+c+d = 50+60+40+30 = 180 DF = 110 / 180 - 90 / 180 DF = 0.11
Odds Radio: Substitute the given values of b,c in the formula, Odds Radio= (a/c)/(b/d) Odds Radio= (50/40)/(60/30)=0.625 %
Confidence Interval: start range = e log(0.625) - (1.96 x √(1/50) + (1/60) + (1/40) + (1/30) ) start range = e log(0.625) - (1.96 x 0.3082207) start range = 0.4456 end range = e log(0.625) + (1.96 x √(1/50) + (1/60) + (1/40) + (1/30) ) end range = e log(0.625) + (1.96 x 0.3082207) end range = 1.4918 Confidence Interval value 0.4456 to 1.4918